Space Data Centers

Full Sector Timeline

~35 milestones from HPE SBC-1 (2017) through Starcloud-1 (Nov 2025) to Google Suncatcher prototypes (2027) and projected cost parity (mid-2030s).
2035 (proj)

Market reaches $39.1B (BIS Research forecast)

67.4% CAGR from $1.78B in 2029; orbital compute captures 20-30% of new DC capacity

BIS Research, GlobeNewsWire

2028 (proj)

Axiom Station free-flying (detaches from ISS)

First commercial free-flying space station

SpaceNews

Early 2027 (proj)

Google Suncatcher prototypes launch with Planet

Two prototype satellites with Trillium TPUs; 15 krad radiation tolerance

Google Research Blog

2027 (proj)

Starcloud-2 launches to sun-synchronous orbit

NVIDIA Blackwell GPUs; Crusoe cloud services; limited commercial capacity from 2027

SpaceNews, DCD

2027 (proj)

Axiom Orbital DC Node on ISS with Kepler/Skyloom/Spacebilt

Kepler 100 Gbps optical links; Phison SSDs; Microchip HPSC processor

Axiom Space

Mar 13, 2026

Starcloud files FCC application for 88,000-satellite constellation

600-850 km sun-synchronous orbits; optical ISL; Ka-band TT&C

SpaceNews

Mar 2026

NVIDIA announces Vera Rubin Space Module

25x H100 compute for space inference; integrated CPU-GPU; purpose-built for orbital DCs

NVIDIA Newsroom, Tom's Hardware

Feb 2026

Axiom secures $350M financing (equity + debt)

Accelerates space station and spacesuit development

Axiom Space

Feb 2026

Sophia Space raises $10M seed

Led by Alpha Funds, KDDI Green Partners; total raised $13.5M

TechCrunch, SpaceWatch

Jan 2026

SpaceX files FCC application for 1M satellite constellation

Orbital data center constellation; 500-2,000 km orbits; accepted for public comment

SpaceNews, FCC

Jan 2026

Blue Origin announces TeraWave constellation

5,408 satellites (5,280 LEO + 128 MEO); 6 Tbps throughput; deployment Q4 2027

Blue Origin, Space.com

Jan 2026

Axiom launches first orbital data center nodes

First ODC nodes to LEO for national security and commercial customers

Introl Blog

Dec 2025

Starcloud-1 trains NanoGPT model in orbit

First AI model trained in space; Gemini/NanoGPT on H100 GPU

CNBC

Dec 2025

K2 Space raises $250M Series C at $3B valuation

Led by Redpoint; T. Rowe Price, Altimeter, Lightspeed participated

PRNewswire, SatNews

Dec 2025

SDA awards $3.5B Tranche 3 contracts

LM $1.1B, L3Harris $843M, Rocket Lab $805M, NG $764M for 72 tracking satellites

SDA.mil, Breaking Defense

Dec 2025

4iG invests $100M in Axiom Space

$30M by end 2025, $70M by March 2026; Hungarian space partnership

Axiom Space

Nov 4, 2025

Google announces Project Suncatcher

81-satellite cluster with TPUs; two prototypes launching early 2027 with Planet Labs

Google Research Blog

Nov 2, 2025

Starcloud-1 launches with NVIDIA H100

First GPU-class compute in orbit; 100x more powerful than prior space computers; 325 km orbit

DCD, CNBC

Nov 2025

IonQ acquires Skyloom (announced)

Optical space communications for quantum networking; $18M Air Force contract included

IonQ, QuantumComputing Report

Oct 28, 2025

FCC votes on Space Modernization NPRM

Proposes orbital debris criteria, identifiability, energy venting, conditional grants

FCC.gov

Jun 2025

Aethero raises $8.4M seed

Seraphim Space involved; NxN edge compute modules with Nvidia Orin NX

TechCrunch

May 14, 2025

China launches Three-Body Computing Constellation (12 sats)

Long March 2D from Jiuquan; 744 TOPS/sat; 5 POPS combined; 2,800 total planned

SpaceNews, CGTN, SCMP

May 2025

Sophia Space raises $3.5M pre-seed

Led by Unlock Ventures; TILE compute module development

SpaceWatch Global

Apr 2025

Aetherflux raises $50M Series A

Led by Index Ventures; a16z, Breakthrough Energy, NEA participated; ~$400M valuation

TechCrunch

Mar 2025

Axiom Space raises $100M

Co-led by 1789 Capital and Type One Ventures at $2B pre-money

GovConWire

Feb 2025

K2 Space raises $110M Series B

Continued scaling of Mega Class satellite manufacturing

SpaceNews

Feb 2025

Starcloud rebrands from Lumen Orbit, raises $10M more

Total ~$24M; NVIDIA Inception Program partnership

GeekWire, DCD

Jan 2025

Loft Orbital raises $170M Series C

Led by Tikehau Capital; BlackRock in earlier round; unicorn status (>$1B valuation)

TechCrunch, Loft Orbital

Dec 2024

Starcloud raises $11M seed round

NFX, Y Combinator, FUSE, Soma Capital, a16z scout, Sequoia scout

GeekWire, DCD

Nov 2024

Cosmic Shielding wins $4M TACFI contract

Pentagon contract to fast-track Plasteel radiation protection; AFWERX/Space Force co-funded

OrbitalToday, SatNews

Oct 23, 2024

ITAR/EAR CSA exception announced

New License Exception for commercial space activities; eases export controls

Federal Register

Oct 2024

Aetherflux founded

Space solar + data center startup; Baiju Bhatt (Robinhood co-founder) $10M pre-seed

TechCrunch

Jul 2024

Aethero + Cosmic Shielding partnership

Plasteel radiation shielding for NxN compute modules; 10x SEE reduction

TechCrunch

Apr 2024

Aethero Deimos satellite launches

First Nvidia Jetson Orin NX in orbit; 100 TOPS edge computing; 1.5U form factor

SpaceNews, Aethero

Feb 2024

K2 Space raises $50M Series A

Mega Class satellite bus development; <$15M per satellite target

PRNewswire

Jan 2024

Starcloud (Lumen Orbit) founded

Founded in Redmond, WA; AI-first orbital data center company; Y Combinator batch

YC, GeekWire

Nov 2023

ESA Zero Debris Charter finalized

Non-binding commitment to Zero Debris by 2030; signed by 12 countries + 200 organizations

ESA

Aug 2023

Axiom Space raises $350M Series C at $2.6B

Largest space station funding round; supports commercial station development

Axiom Space

Jun 2023

NASA TBIRD achieves 200 Gbps optical downlink

4.8 TB error-free data in 5 min; fastest space-to-ground laser comms; MIT Lincoln Lab

NASA Goddard

Apr 13, 2023

Kepler Communications raises $92M Series C

IA Ventures led; completing optical data relay constellation; 100 Gbps WARP terminal

Kepler, SpaceNews

Sep 29, 2022

FCC adopts 5-year deorbit rule (FCC 22-74)

Replaces 25-year guideline; applies to all new satellite licenses; effective Sep 2024

FCC.gov

Feb 20, 2021

HPE Spaceborne Computer-2 launches to ISS

Northrop Grumman NG-15 resupply; 130 TB storage; edge computing demos with Microsoft Azure

ISS National Lab

Aug 14, 2017

HPE Spaceborne Computer-1 launches to ISS

First commercial supercomputer in space; SpaceX CRS-12; ran 615 days with zero unrecoverable errors

ISS National Lab

Mid-2030s (proj)

Projected cost parity: orbital vs terrestrial DCs

Requires Starship at ~$200/kg and 180 launches/year; orbital LCOE matches terrestrial

Google Research, Deutsche Bank

44 of 44 events shown. Sources: company announcements, FCC filings, press releases.

Regulatory Landscape

FCC 5-year deorbit rule, FCC Space Modernization NPRM, ITAR/EAR CSA exception, ESA Zero Debris Charter, NOAA licensing, and ITU spectrum congestion.
Regulation / PolicyAgency / BodyDateKey ProvisionsImpact on Space DCsSource
5-Year Deorbit Rule (FCC 22-74)FCC (US)Adopted Sep 29, 2022; Effective Sep 29, 2024All new satellites must deorbit within 5 years of end-of-life (replaces 25-year guideline)Limits useful satellite lifetime; increases replacement cadence and cost; favors distributed architecturesFCC.gov
Space Modernization NPRMFCC (US)Published Oct 7, 2025Proposes orbital debris criteria, satellite identifiability, energy venting requirements, conditional grantsCould add compliance costs; standardizes space DC operations; may require new design featuresFCC.gov
ITAR/EAR License Exception CSAFCC / State Dept (US)Announced Oct 23, 2024New export control exception for commercial space activities (Lunar Gateway, Mars, CLEO)Eases international partnerships for commercial space computing; reduces export frictionFederal Register
ESA Zero Debris CharterESA (Europe)Finalized Nov 2023; Target 2030Non-binding commitment to achieve zero new debris by 2030; signed by 12 countries + 200 organizationsMay influence European orbital DC designs; voluntary compliance standardESA
NOAA Remote Sensing LicensingNOAA (US)Ongoing / EvolvingLicensing requirements for Earth observation from space; evolving to cover compute-enhanced sensingSpace-based computing with EO may trigger new licensing requirementsNOAA
ITU Spectrum CoordinationITU (International)OngoingCoordination for inter-satellite links and ground station frequencies; increasing congestionGrowing competition for Ka-band, optical, and other spectrum for space DC communicationsITU
Golden Dome AuthorizationUS Congress / DoDFY2026 Budget$13.4B FY2026 for integrated space and ground defense layerCreates major government demand signal for orbital computing capabilitiesAerospace CSPS, SatNews
SDA Proliferated ArchitectureSDA / Space Force2020-ongoingHundreds of small satellites in LEO mesh for missile tracking and data relayEstablishes military demand and standards for space-based computing and networkingSDA.mil

Regulatory landscape for space-based data centers. Sources: government agencies, regulatory filings.

Risks & Challenges

7 key risks quantified: radiation, thermal management (834K m² radiators for 1GW), launch failure, space debris (300K Starlink maneuvers/yr), latency, serviceability, and regulatory uncertainty.
Radiation DamageHigh

Google Trillium TPU: HBM fails at ~2 krad; ≥15 krad for core logic; COTS GPUs not rad-rated

Plasteel shielding (10x SEE reduction); software hardening (HPE approach); traditional rad-hard (5-10x cost)

Google Research, Cosmic Shielding, HPE

Thermal ManagementVery High

1 GW DC needs 834,000 m² radiators (~83 hectares); ISS only handles 70 kW with 422 m²

Distributed small sats; liquid droplet radiators (10x lighter); advanced heat pipe technology

Medium Analysis, NASA EATCS

Launch Failure / CostHigh

No mature insurance market; only 300 of 10,000+ sats insured; Falcon 9 at $2,600/kg today

Starship target $13-20/kg; constellation redundancy; in-orbit spares

DCD, NextBigFuture

Space Debris / CollisionHigh

40,230 tracked objects; 300K Starlink maneuvers/year; 500-800 km band critically crowded

5-year deorbit rule; active debris removal; lower orbit (higher drag = faster deorbit)

ESA 2025, Space4Peace, FCC

LatencyMedium

LEO: 20-40 ms round-trip vs <1 ms fiber; GEO: 480-600 ms

LEO for lowest latency; mesh networking; edge caching in orbit

Physics, industry standard

ServiceabilityVery High

Zero physical access once deployed; no repair, upgrade, or component replacement possible

Design for reliability; modular constellation (replace entire sats); autonomous fault recovery

Gartner, IEEE Spectrum

Regulatory UncertaintyMedium

FCC Space Modernization NPRM pending; ITAR/EAR complexity; ITU spectrum congestion growing

Engage regulators early; comply with strictest standards; diversify across jurisdictions

FCC, Federal Register

Power ScalingHigh

Space solar panels cost 1,000x terrestrial; ISS: 120 kW total; 1 GW needs massive arrays

Thin-film solar (150-250 W/kg target); space solar beamed power (Aetherflux approach)

IEEE Spectrum, ScienceDirect

Bandwidth BottleneckMedium

Ground contact 5-15 min/pass in LEO; 200 Gbps max demonstrated (TBIRD)

Optical ISL mesh (100 Gbps-1.6 Tbps); in-orbit processing (reduce downlink 30,000x)

NASA TBIRD, HPE, Google

Technology MaturityMedium-High

First GPU in space only Nov 2025; no multi-year space DC operational data yet

Phased approach (crawl/walk/run); ISS heritage (HPE 615 days); incremental scaling

Deutsche Bank, industry consensus

Risk assessment for space-based data center development. Sources: industry reports, NASA, ESA, regulatory filings.

Space Debris Trend

Tracked objects grew from ~25K to ~40K (2020-2025). Starlink collision avoidance maneuvers tripled year-over-year to 300K in 2025.
MetricValueYear / PeriodSource
Tracked Objects
Total tracked objects (>10 cm)~20,000Oct 2019Historical baseline
Total tracked objects (>10 cm)~31,000Feb 2024NASA SVS
Total tracked objects (>10 cm)40,230Apr 2025ESA SDUP
Collision Avoidance
Starlink collision avoidance maneuvers50,666Jun-Dec 2023 (6 months)Space.com
Starlink collision avoidance maneuvers144,404Dec 2024-May 2025 (6 months)Space Intel Report
Starlink collision avoidance maneuvers~300,000Full year 2025Space4Peace
Total Starlink satellites in orbit~7,000+End 2025SpaceX filings
Estimated Untracked
Estimated objects 1-10 cm~1,200,0002025ESA Space Environment Report
Estimated objects 1 mm - 1 cm~140,000,0002025ESA Space Environment Report
Other
Large debris fragments (>10 cm)37,000+Jun 2023ESA/IADC reports
Active payloads in orbit~11,0002025ESA SDUP
Non-deliberate fragmentation events/year10.5 per year2024 averageESA 2025 Report
New catalogued fragments (2024)3,000+2024ESA 2025 Report
Peak debris density altitude500-800 kmCurrentESA analysis
Average LEO impact velocity10 km/sPhysicsOrbital mechanics
Maximum LEO impact velocity>14 km/sPhysicsOrbital mechanics
Satellites insured (of 10,000+ total)~3002025DCD reporting
FCC deorbit rule5 years post-end-of-lifeEffective Sep 29, 2024FCC 22-74
ESA Zero Debris targetZero new debris by 2030Charter signed Nov 2023ESA

Space debris environment data relevant to orbital data center planning. Sources: ESA, NASA, SpaceX filings, FCC.

Bull vs Bear Case

Morgan Stanley and Musk see transformative potential. Gartner calls it 'peak insanity.' IEEE estimates 1 GW orbital DC at $51B vs $16B terrestrial.
Bull

Morgan Stanley: orbital DCs explain SpaceX valuation doubling in 6 months; Musk: 'In 36 months but probably closer to 30 months' (WEF Jan 2026); BIS Research: $39.1B by 2035 at 67.4% CAGR

Market Timing
Bear

Gartner (Bill Ray): 'Peak insanity — datacenters in space won't analyze data on Earth for decades, if ever'; Sam Altman: 'honestly ridiculous' with current landscape; Deutsche Bank: 'well into the 2030s before parity'

Key Inflection Point: First commercial workload delivering revenue (est. 2027-2028)

Bull

Google: 8x solar advantage (confirmed in research paper); Starcloud: $0.002/kWh projected vs $0.045/kWh terrestrial US grid; 95% capacity factor vs 23.5% Earth; zero cooling energy cost (PUE = 1.0 vs 1.56 avg)

Energy Economics
Bear

Space solar panels: ~$200/W (100x terrestrial $2/W); McCalip: orbital LCOE $891/MWh vs terrestrial $398/MWh; even with 8x productivity, cost gap dominates; Starcloud 22x claim is theoretical max — 10x more realistic

Key Inflection Point: Starship at $200/kg (Google's target) closes energy LCOE gap by mid-2030s

Bull

Space ambient at 2.7K (-270°C); zero water usage (vs 17B gal/yr US DCs); zero HVAC infrastructure; passive radiative cooling eliminates 30-40% of terrestrial OpEx

Cooling Physics
Bear

Stefan-Boltzmann law: ISS rejects only 70 kW via 422 m² radiators (166 W/m²); 1 GW DC needs 834,000 m² (83 hectares) of radiators weighing 2,250 tonnes; physics expert: 'Never happening!' at current tech

Key Inflection Point: Liquid Droplet Radiators (10x lighter) or active space-rated heat pumps proven at scale (est. 2028+)

Bull

Falcon 9 internal: $629/kg; Starship 6-flight reuse: $78-94/kg; 50-flight reuse: $13-20/kg; Google needs only $200/kg for parity — 10x reduction from today

Launch Cost Trajectory
Bear

Starship unproven at scale; current customer price still $2,600/kg; high reuse (50+ flights) never demonstrated; 180 launches/year required for Google model — F9 record is ~100/yr

Key Inflection Point: Starship achieves 20+ flights per booster with 100+ launches/year cadence (est. 2028-2030)

Bull

US DC power demand +22% in 2025, tripling to 150 GW by 2030; 10,300 projects / 1,400 GW in grid queue; 36 projects / $162B blocked or delayed; 3-7 year grid connection wait times; utilities cannot deliver GW in 12-24 months

Terrestrial Grid Crisis
Bear

Grid crisis may be temporary — nuclear SMRs, new transmission, distributed solar deploying; AI efficiency gains (smaller models, quantization) reduce power per FLOP; DC industry pivoting to self-generation (natural gas, on-site solar)

Key Inflection Point: Persistent 5+ year grid queues making orbital faster-to-deploy than terrestrial (requires Starship at scale)

Bull

SDA Tranche 3: $3.5B for 72 tracking satellites with on-orbit data processing; Golden Dome: $13.4B FY2026; K2 STRATFI: $60M; Space Force MILNET: 480-sat military constellation; defense demand independent of commercial ROI

Defense / Government Pull
Bear

Military demand alone won't scale commercial market; classified programs don't create open ecosystem; defense-grade requirements increase costs 5-10x; no confirmed '$500M orbital computing budget' in public documents

Key Inflection Point: Defense contracts explicitly specify 'on-orbit AI compute' as requirement (beginning with SDA Tranche 3)

Bull

Hyperscalers spending $443B in 2025, $602B in 2026; NVIDIA Blackwell sold out; Vera Rubin Space Module delivers 25x H100 for orbit; demand growing faster than terrestrial capacity can deploy

AI Compute Demand
Bear

Not all AI workloads suit orbital deployment; training requires tight GPU interconnect (100 Gbps ISL << NVLink 900 GB/s); inference is viable but smaller addressable market; most AI workload growth served by terrestrial

Key Inflection Point: Orbital inference becomes competitive for specific use cases (EO, edge, defense) even at premium pricing

Bull

HPE ran 615 days on ISS with zero unrecoverable compute errors; HPE SBC-2 achieved 30,000x data reduction (2.8 GB → 92 KB, verified); Starcloud-1 trained NanoGPT + ran Gemma in orbit; Aethero Deimos: 100 TOPS in orbit (Aug 2024)

Technical Feasibility
Bear

9% GPU failure rate in orbital conditions (Meta study); radiation at 4nm node worsens SEE vulnerability; no multi-year commercial space DC operational data exists; HPE ISS environment is heavily shielded — LEO free-flight is harsher

Key Inflection Point: Multi-year commercial workload uptime demonstrated on free-flying satellite (not ISS-hosted)

Bull

Constellation architecture provides redundancy — individual satellite loss is tolerable; operating cost per satellite is declining; no single point of failure in distributed mesh

Insurance & Risk
Bear

Only ~300 of 12,787 satellites insured; <50 of 9,000+ LEO sats insured; insurers lost $500M+ in 2023; major carriers (Allianz, AIG, Swiss Re) withdrawing; LEO policy costs $500K-1M vs $200-300M for GEO

Key Inflection Point: Space DC insurance products emerge with actuarial data from constellation operations (est. 2028+)

Bull

FCC 5-year deorbit rule limits long-term debris growth; distributed small sats deorbit faster from low altitudes; maneuverability improving (300K Starlink avoidance maneuvers in 2025, zero collisions)

Space Debris
Bear

40,230 tracked objects (Apr 2025); 520-1,000 km at 'potential runaway threshold' for Kessler syndrome; Feb 9, 2026: 441 conjunctions in single day with >11 km/s velocities; adding 1M+ satellites dramatically increases collision probability

Key Inflection Point: Active debris removal commercially operational; or sustained zero-collision record at mega-constellation scale

Bull

McCalip model shows gap narrows dramatically if satellite costs halved: from 3x to near-parity; zero land cost, zero water, zero grid connection fees — terrestrial hidden costs rising

Cost Comparison (1 GW)
Bear

McCalip baseline: $42.4B orbital vs $14.8B terrestrial (2.9x gap); IEEE Spectrum: $51B for 1 GW orbital; 4,300 satellites needed; only SpaceX positioned to attempt

Key Inflection Point: Satellite manufacturing at Starlink-like volume + cost drives per-sat cost below McCalip sensitivity threshold

Bull

Zero land use, zero water, zero grid emissions; addresses DC sustainability crisis; carbon-free solar 24/7; Fortune: Google CEO Pichai cites space DCs as 'environment' solution

Environmental Impact
Bear

Rocket launches emit CO2 + soot; decommissioned hardware = high-tech space junk; ASCEND study: needs launcher '10x less emissive' than current rockets; alumina from reentry damages ozone layer

Key Inflection Point: Methane/LOX rockets + reusability reduce per-payload emissions to negligible levels

Bull

Starlink V3: 1 Tbps per satellite; Google bench: 1.6 Tbps bidirectional optical; NASA TBIRD: 200 Gbps ground link; Blue Origin TeraWave: 6 Tbps network aggregate

Bandwidth / Connectivity
Bear

Current operational ISL: 100 Gbps (vs NVLink 900 GB/s in terrestrial GPU clusters — 9,000x gap); inter-satellite coordination for distributed training is unsolved at required bandwidth; ground contact 5-15 min/pass in LEO

Key Inflection Point: 1 Tbps+ ISL sustained in production (not bench) closes gap for inference; training may always require terrestrial interconnect

Bull

Starlink proved mega-constellation skeptics wrong: $1.4B (2022) → $6.6B revenue (2024); 2.7M subscribers across 75 countries; industry dismissed LEO broadband in 2015-2018

Historical Analogy
Bear

Starlink ≠ orbital DC: (1) clear ROI from underserved broadband market, (2) connectivity is simpler than compute, (3) different physics constraints (power/cooling/failure), (4) terrestrial DCs already efficient (unlike rural broadband)

Key Inflection Point: First profitable commercial orbital compute workload — even at small scale — breaks the analogy objection

Bull

Near-term: satellite imagery processing (Planet 30 TB/day), defense ISR, edge inference, autonomous satellite ops, data compression/filtering; all 'data produced in space, consumed in space' (acknowledged even by Gartner)

Viable Workloads
Bear

Gartner: only 'data produced in space for consumption in space' makes sense; general terrestrial AI training/inference in orbit 'won't happen for decades, if ever'; financial trading and <1ms workloads impossible from orbit

Key Inflection Point: Demonstrated cost advantage for at least one specific workload category (likely EO processing, est. 2027-2028)

Critical Unknowns

  1. 1.Starship reuse rate: 50+ flights/booster enables $13-20/kg → closes energy LCOE gap (NextBigFuture)
  2. 2.GPU failure rates in LEO radiation: 9% (Meta study) may be optimistic or pessimistic — no long-term data (Per Aspera)
  3. 3.Thermal solution maturity: Liquid Droplet Radiators or active heat pumps proven at >100 kW scale (Wikipedia LDR)
  4. 4.Terrestrial grid crisis duration: If 5+ year queues persist, orbital becomes faster-to-deploy (ENR, LandGate)
  5. 5.Defense budget willingness: If DoD subsidizes orbital compute for strategic advantage, commercial economics less relevant (SDA.mil)
  6. 6.Satellite manufacturing cost: If mass-produced at Starlink volume, McCalip model approaches parity (McCalip Calculator)
  7. 7.Timeline consensus: Prototypes 2027-28; pilot constellation 2028-30; cost parity earliest mid-2030s (Deutsche Bank, Google)
Space Data Centers - Risks | Sterling